In our latest article, we discussed the difference between CPM and eCPM. Now let’s dig more into details why eCPM can be misleading.
Publishers often evaluate the overall results by looking at the average eCPM. This can be right for a quick evaluation of overall performance and comparison. However, it doesn’t describe the full picture. Unless all other factors are similar, comparing the average eCPMs can be misleading. Let us explain why with an example.
We took the results from one particular placement over a short period. We analyzed how different bid values are allocated and where does the average eCPM result stand. The results were as followed:
- 268 731 impressions;
- 98,86% fill rate;
- average eCPM – € 0,75;
- total revenues – € 202,45
How to calculate average eCPM?
Average eCPM = (Total Revenues / Total Impressions) x 1000
(€ 202,45 / 268 731) x 1000 = € 0,75 (average eCPM)
We divided all impressions into deciles, each decile having a similar amount of impressions with particular bid prices. The last decile has a little fewer impressions, but that does not affect the findings.
What can we learn from this data?
The changing part is the revenues. The graph above shows how much revenues are generated at different price deciles. Naturally, there were more impressions with the lower prices. Therefore the intervals are shorter on the first deciles and become wider and wider as the price increases.
We have to pay attention to how much of overall revenues each of these deciles contribute.
The last decile (bid price range from €2,31 – even €100) has the most extended CPM interval because a minimal amount of impressions falls into this top tier. Usually, these are retargeting campaigns, where advertisers are willing to pay a high price for a single impression. As we can see this price level contributes to the most significant part (45,7%) of the overall ad revenues.
Only 11,5 % of revenues we get from the impressions sold around the average price range.
All deciles with a smaller than the average bid price contribute 22,3% to final revenues.
Among these, the bids from €0,1 – €0,6 generate only 1% of all revenues. These low CPMs usually come from less developed markets where programmatic is not as advanced as in the rest of the world. These could also be fresh cookies that advertisers know very little about and are not willing to place higher bids. Even though the low CPMs bring down the average price, it is still a significant revenue source. Quite often the low bids are used by advertisers for probing the inventory, and restricting them with a floor price can result in advertisers not bidding higher bids later.
Stigma around ‘average results’
We usually compare and make decisions based on the average ecpm, but in the end, it does not represent the whole picture accurately.
In between the CPMs that bring the biggest and lowest revenues is the ‘average price’ which is the one that everyone takes notice of although it contributes to the lowest overall revenue shares. As yield managers at Setupad, we see our final result is evaluated based on the average result. Of course this is mathematically correct, however, in this case, it does a lousy job of reflecting all the revenue sources. The average price is disregarding the most valuable vital contributors where yield management is making the most significant impact.
Analyzing these results, we can conclude that the most significant revenue shares come from the highest CPMs, even though these come in a significantly lesser amount than impressions with lower CPMs.
The skill to optimize the highest CPMs determines the final results as they are the biggest contributors to the final amount of revenues. Selling more impressions for €0,5 or €0,6 CPM will not mean a whole lot when looking at the big picture. However, it is essential to understand that the best practice is to monetize as much as possible of the entire ad inventory in order not to lose a significant part of revenues. Optimization is key to getting higher revenues, and this should be the primary focus when it comes to selling ad inventory.